ESTRO 2023

Session Item

Monday

May 15

10:30 - 11:30

Strauss 1

Novel applications of MR imaging

Chair: , Germany;

Co-Chair: , The Netherlands

Session Code: 3220

Session Type: Proffered Papers

Track: Physics

11:20 - 11:30

Histopathology-validated GTV delineations of intra-prostatic lesions with mpMRI and PSMA-PET

Josefine Grefve, Sweden

Presentation Number: OC-0781

Abstract

Abstract Title:

Histopathology-validated GTV delineations of intra-prostatic lesions with mpMRI and PSMA-PET

Authors:

Josefine Grefve¹, Kristina Sandgren¹, Joakim Jonsson¹, Angsana Keeratijarut Lindberg¹, Erik Nilsson¹, Sara Strandberg², Anders Bergh³, Karin Söderkvist⁴, Björn Zackrisson⁴, Adalsteinn Gunnlaugsson⁵, Mathieu Moreau⁶, Camilla Thellenberg Karlsson⁴, Lars E. Olsson⁷, Bengt Friedrich⁸, Anders Widmark⁴, Lennart Blomqvist⁹, Vibeke Berg Loegager¹⁰, Jan Axelsson¹, Mattias Ögren², Margareta Ögren², Katrine Riklund¹¹, Tufve Nyholm¹

¹Umea University, Department of Radiation Sciences, Radiation Physics, Umea, Sweden; ²Umea University, Department of Radiation Sciences, Diagnostic Radiology, Umea, Sweden; ³Umea University, Department of Medical Biosciences, Pathology, Umea, Sweden; ⁴Umea University, Department of Radiation Sciences, Oncology, Umea, Sweden; ⁵Skane University Hospital, Department of Hematology, Oncology and Radiation Physics, Lund, Sweden; ⁶Skane University Hospital, Department of Hematology, Oncology and Radiation Physics, Lund , Sweden; ⁷Lund University, Department of Translational Medicine, Medical Radiation Physics, Malmo, Sweden; ⁸Umea University, Department of Surgical and Perioperative Sciences, Urology and Andrology, Umea, Sweden; ⁹Karolinska Institute, Department of Molecular Medicine and Surgery, Solna, Sweden; ¹⁰Copenhagen University Hospital in Herlev, Department of Radiology, Herlev, Denmark; ¹¹Umea University, Department of Radiation Sciences, Diagnostic Radiology, Umea , Sweden

Show Affiliations

Purpose or Objective

There is growing evidence that focal boost to intra-prostatic lesions is beneficial for patients with high-risk prostate cancer. Delineation of boost regions can be done using MRI with T2w imaging, diffusion-weighted imaging (DWI) or perfusion (dynamic contrast-enhanced, DCE) imaging, or with PET, where PSMA currently is one of the dominant radiotracers for prostate cancer. The delineation guided by the different modalities will result in a variation in gross tumor volumes (GTV), which may impact the effectiveness of the treatment. In the present study we investigate the characteristics of manual intra-prostatic GTV delineations based on different MR sequences and PSMA-PET using whole-mount histopathology as gold standard.

Material and Methods

15 prostate cancer patients with high risk profile (Gleason >=8) were imaged with PSMA-PET/MR prior to surgery. The prostate specimens were placed in individualized prostate molds with slits (5 mm separation) corresponding to the imaging planes for the histopathology slicing. Gleason grade regions 3, 4 and 5 were delineated in each slice and through a careful registration procedure transferred to the coordinate system of the PET/MR examination. Four radiation oncologists individually delineated GTVs, based on T2w-MRI, diffusion-MRI (DW-MRI), DCE-MRI, and PSMA-PET, respectively. The Staple algorithm was used to combine the delineations for each modality.

Results

All results are presented for the combined (Staple algorithm) GTV delineation for each imaging modality. In regions with Gleason grade 4 (15 patients) the mean volumetric overlap with standard deviation was 38% ± 23% (T2w-MRI), 37% ± 22% (DW-MRI), 34% ± 20% (DCE-MRI) and 51% ± 27% (PSMA-PET). For regions with Gleason grade 5 (3 patients) the overlap was 36% ± 36% (T2w-MRI), 69% ± 29% (DW-MRI), 37% ± 52% (DCE-MRI) and 71% ± 17% (PSMA-PET). For regions with Gleason grade 3 (14 patients) the overlap was 25% ± 24% (T2w-MRI), 26% ± 24% (DW-MRI), 18% ± 21% (DCE-MRI) and 25% ± 31% (PSMA-PET). Including regions with all Gleason grades the overlap was 35% ± 24% (T2w-MRI), 37% ± 24% (DW-MRI), 31% ± 18% (DCE-MRI) and 42% ± 27% (PSMA-PET).

A combination of the different modalities yielded a mean volumetric overlap (all Gleason grades) of 51% ± 25% for the combination of T2w-MRI + DW-MRI; by also including DCE-MRI it became 55% ± 24% and by combining all four modalities the overlap was 60% ± 25%.

The mean Dice score (all Gleason grades) with standard deviation was 38% ± 23% (T2w-MRI), 39% ± 22% (DW-MRI), 36% ± 19% (DCE-MRI) and 38% ± 20% (PSMA-PET).

Figure 1 a) Histopathology slice with delineated Gleason grade regions, b) PSMA-PET uptake, c) GTV delineations from the four different observers and d) the combined GTV from the Staple algorithm.

Conclusion

The largest overlap between ground truth and the delineated lesions was found for higher Gleason grades (>=4) using PSMA-PET or combinations of modalities. The variation in overlap between patients was significant and warrants further analysis.